venturebeat.com

24 articles in the Truth Foundry index. Each links out to the original.

• OpenAI Launches ChatGPT Work to Transform AI into Workplace Platform · OpenAI introduces ChatGPT Work, a cloud-based AI agent designed to manage complex tasks across email, Slack, and calendars, aiming to reposition ChatGPT as a workplace platform.
• Enterprises Deploy AI Agents Prematurely, Leaving GPUs Underutilized · 86% of enterprises report GPU utilization at 50% or less while deploying AI agents without adequate controls.
• Enterprise AI Faces Evaluation Gap as Autonomy Outpaces Verification · Enterprise AI agents are gaining autonomy faster than companies can verify their reliability, leading to customer-facing failures despite passing internal tests.
• Red Hat's Brian Gracely Warns of Cost, Security, and Culture Hurdles in Enterprise AI Agents · Red Hat's Brian Gracely outlines critical cost, security, and cultural barriers preventing enterprise AI agents from scaling beyond pilot programs.
• Box Report: Enterprise AI Leaders Outperform Peers via Content Governance · Box survey reveals content access and governance are key differentiators for high-ROI enterprise AI.
• Anthropic Expands Claude Cowork to Mobile and Web Amid Non-Developer Usage Surge · Anthropic launches Claude Cowork on mobile and web, revealing most users utilize it for administrative tasks rather than coding.
• MongoDB Atlas Powers Digital-Native AI Startups to Ditch Rigid Databases · Three startups leverage MongoDB Atlas to build flexible, agent-native data stacks.
• Trunk Tools Cuts Document Review Time from 60 Days to 10 with Specialized AI Stack · Construction firm Trunk Tools reduced review cycles by 83% using a custom three-layer AI architecture.
• Two-Thirds of Enterprises Hedge AI Strategy After Claude Fable 5 Blackout · Majority of enterprises now blend closed and open-weight AI models to avoid vendor lock-in and government export restrictions.
• Alibaba's SkillWeaver Cuts AI Agent Token Use by 99% via Smart Tool Routing · Alibaba's SkillWeaver framework reduces AI agent token consumption by 99% through iterative skill-aware decomposition.
• Morgan Stanley Cuts P&L Reconciliation Work in Half by Reducing Agent Autonomy · Morgan Stanley reduced P&L reconciliation time by 50% by keeping humans tightly in the loop with AI agents.
• Couchbase Unveils AI Data Plane for Enterprise Agent Memory and Edge Context · Couchbase launches AI Data Plane to provide persistent agent memory and real-time context retrieval across cloud and edge environments.
• Meituan Unveils LongCat-2.0: 1.6T Chinese Chip AI Model Disrupts OpenRouter · Meituan releases LongCat-2.0, a 1.6T parameter agentic coding model trained on domestic Chinese chips.
• DeepSeek Open Sources DSpark Framework for 85% Faster LLM Inference · DeepSeek releases MIT-licensed DSpark framework to accelerate large language model inference speeds by up to 85%.
• Prompt Injection Attacks Target Enterprise AI Agents and RAG Pipelines · Cybercriminals exploit LLM vulnerabilities to hijack agents and steal credentials via prompt injection attacks.
• Krea 2 Raw and Turbo Models Released as Open Weights for Enterprise AI · Krea launches Krea 2 Raw and Turbo models with 2-second generation speed under a custom enterprise license.
• Anthropic Launches Claude Tag, Replacing Slack App with Persistent AI Teammate · Anthropic replaces its Slack app with Claude Tag, a persistent AI agent for enterprise teams.
• Alibaba's HappyHorse 1.1 AI Video Model Ranks No. 2 Globally Amid Competitor Setbacks · Alibaba Cloud launches HappyHorse 1.1, a production-ready AI video model, ranking second globally as Sora and Seedance withdraw.
• AI Agents Learn to Rewrite Their Own Rules, Boosting Performance by 60% · Researchers introduce Self-Harness, a framework allowing AI agents to autonomously optimize their operating rules.
• 7,000 Langflow Servers Exposed to RCE via Path Traversal Vulnerability · Path traversal flaw in Langflow allows unauthenticated remote code execution on thousands of servers.
• Stanford's DeLM Cuts Multi-Agent AI Costs 50% Without Central Controller · Stanford's DeLM framework reduces multi-agent AI costs by 50% through decentralized coordination without a central orchestrator.
• PixelRAG System Beats Text Parsers on Accuracy and Cuts AI Agent Token Costs 10x · New PixelRAG system skips text parsing to improve RAG accuracy and slash AI agent token usage by 10x
• Anthropic CEO Advocates for FAA-Style AI Regulations to Ensure Safety and Compliance · Anthropic CEO Dario Amodei calls for FAA-style regulations for AI models to ensure safety and address labor displacement.
• MassMutual Adopts 12-Month AI Contracts to Maintain Flexibility and Productivity · MassMutual uses short-term vendor contracts to maintain optionality while achieving 30% productivity gains.